def calculate_roi_perf(self, roi_name):
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
real_tac = np.array(self.rois.get_concentrations(roi_name))
real_tac -= real_tac[0]
times = self.rois.get_time_list()
print 'times', times
new_times = np.arange(times[0], times[-1] + 1, 1)
print map(int, times)
smoothed_tac, in_tac = express.spline_interpolation(
input_tac, times, new_times)
s_tissue_tac, s_tissue_tac_pars = express.spline_interpolation(
real_tac, times, new_times, ss=0.1)
modeled_tac, modeled_pars = express.calc_tissue_tac_mrx_conv(
smoothed_tac,
time_steps=new_times,
time_subset=times,
rc_type='lognorm',
mtts=np.arange(1, 70, 4),
bvs=np.arange(0, 1, 0.2),
rc_sigma=np.arange(0.01, 1, 0.2))
# modeled_tac, modeled_pars = express.calc_tissue_tac_mrx(in_tac, times=times, mtts=np.arange(5,80,1), bvs=np.arange(0.1,0.9,0.1))
modeled_pars = np.array(modeled_pars)
print modeled_pars.shape
ssd = np.sum((modeled_tac - real_tac)**2, axis=-1)
best_choice = modeled_tac[ssd.argmin()]
subset_array = ssd.argsort()[:10]
print modeled_pars[ssd.argmin()]
plt.subplot(3, 1, 1)
plt.plot(times, real_tac, 'o-')
plt.plot(times, best_choice, '--')
gr = np.array([modeled_tac[i] for i in subset_array])
# print gr.shape
plt.plot(times, gr.T, '-', alpha=0.3)
plt.subplot(3, 1, 2)
bv_subset = modeled_pars[:, 1][subset_array]
cmap = plt.cm.coolwarm(bv_subset / np.max(bv_subset))
print modeled_pars[subset_array]
plt.scatter(modeled_pars[:, 0][subset_array],
modeled_pars[:, 1][subset_array],
marker='o',
edgecolors=cmap)
plt.subplot(3, 1, 3)
plt.plot(modeled_pars[:, 0][subset_array], ssd[subset_array], 'o')
plt.show()
def make_4dvol_frommaps(self):
#load maps bv,mtt,bf,sigma
perfmap_art = nib.load(
self.dir_manager.get_path('perf_map')).get_data()
perfmap_port = nib.load(
os.path.join(
self.dir_manager.get_path('perf_map')[:-7] +
'2.nii.gz')).get_data()
perfmap = np.concatenate(
(perfmap_art[..., (1, 0, 3)], perfmap_port[..., (1, 0, 3)]),
axis=3)
perfmap[..., (0, 3)] /= 100.
perfmap2d = perfmap.reshape(
(np.prod(perfmap.shape[:-1]), perfmap.shape[-1]))
pars_subset = perfmap2d[np.where(perfmap2d[..., 0] != 0)]
#load input tacs
times = self.rois.get_time_list()
new_time_steps = np.arange(times[0], times[-1] + 1, 2)
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
input_tac_smoothed, input_tac_splines = express.spline_interpolation(
input_tac, times, new_time_steps)
try:
input_tac2 = np.array(self.rois.get_concentrations('porta'))
input_tac2 -= input_tac2[0]
input_tac_smoothed2, input_tac_splines2 = express.spline_interpolation(
input_tac2, times, new_time_steps)
except:
input_tac2 = None
pass
vol4d = express.calc_tissue_tac_from_pars(input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=[60, 70, 80],
params=pars_subset)
vol4d = vol4d.reshape(
(np.prod(perfmap.shape[:-1]), vol4d_2d.shape[-1]))
hdr = nib.load(self.dir_manager.get_path('4d')).get_header()
mrx = hdr.get_sform()
nii_im = nib.Nifti1Image(out_vol, mrx, hdr)
nib.save(
nii_im,
os.path.join(self.dir_manager.get_path('4D')[:-4] + '2.nii.gz'))
def make_4dvol_frommaps(self):
#load maps bv,mtt,bf,sigma
perfmap_art = nib.load(self.dir_manager.get_path('perf_map')).get_data()
perfmap_port = nib.load(os.path.join(self.dir_manager.get_path('perf_map')[:-7]+'2.nii.gz')).get_data()
perfmap = np.concatenate((perfmap_art[...,(1,0,3)],perfmap_port[...,(1,0,3)]),axis=3)
perfmap[...,(0,3)] /= 100.
perfmap2d = perfmap.reshape( (np.prod(perfmap.shape[:-1]),perfmap.shape[-1]) )
pars_subset = perfmap2d[ np.where(perfmap2d[...,0] != 0)]
#load input tacs
times = self.rois.get_time_list()
new_time_steps = np.arange(times[0], times[-1] + 1, 2)
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
input_tac_smoothed, input_tac_splines = express.spline_interpolation(input_tac, times, new_time_steps)
try:
input_tac2 = np.array(self.rois.get_concentrations('porta'))
input_tac2 -= input_tac2[0]
input_tac_smoothed2, input_tac_splines2 = express.spline_interpolation(input_tac2, times, new_time_steps)
except:
input_tac2 = None
pass
vol4d = express.calc_tissue_tac_from_pars(input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=[60,70,80],
params=pars_subset)
vol4d = vol4d.reshape( (np.prod(perfmap.shape[:-1]),vol4d_2d.shape[-1]) )
hdr = nib.load(self.dir_manager.get_path('4d')).get_header()
mrx = hdr.get_sform()
nii_im = nib.Nifti1Image(out_vol, mrx, hdr)
nib.save(nii_im, os.path.join(self.dir_manager.get_path('4D')[:-4]+'2.nii.gz'))
def calculate_roi_perf(self):
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
real_tac = np.array(self.rois.get_concentrations('roi1'))
print real_tac
real_tac -= real_tac[0]
times = self.rois.get_time_list()
print 'times', times
new_times = np.arange(times[0], times[-1] + 1, 1)
print map(int, times)
smoothed_tac, in_tac = express.spline_interpolation(input_tac, times, new_times)
s_tissue_tac, s_tissue_tac_pars = express.spline_interpolation(real_tac, times, new_times, ss=0.1)
modeled_tac, modeled_pars = express.calc_tissue_tac_mrx_conv(smoothed_tac,
time_steps=new_times,
time_subset=times,
rc_type='lognorm',
mtts=np.arange(1, 70, 4),
bvs=np.arange(0, 1, 0.1),
rc_sigma=np.arange(0.01, 1, 0.2))
# modeled_tac, modeled_pars = express.calc_tissue_tac_mrx(in_tac, times=times, mtts=np.arange(5,80,1), bvs=np.arange(0.1,0.9,0.1))
modeled_pars = np.array(modeled_pars)
print modeled_pars.shape
ssd = np.sum((modeled_tac - real_tac) ** 2, axis=-1)
best_choice = modeled_tac[ssd.argmin()]
subset_array = ssd.argsort()[:10]
print modeled_pars[ssd.argmin()]
plt.subplot(3, 1, 1)
plt.plot(times, real_tac, 'o-')
plt.plot(times, best_choice, '--')
gr = np.array([modeled_tac[i] for i in subset_array])
#print gr.shape
plt.plot(times, gr.T, '-', alpha=0.3)
plt.subplot(3, 1, 2)
bv_subset = modeled_pars[:, 1][subset_array]
cmap = plt.cm.coolwarm(bv_subset / np.max(bv_subset))
print modeled_pars[subset_array]
plt.scatter(modeled_pars[:, 0][subset_array], modeled_pars[:, 1][subset_array], marker='o', edgecolors=cmap)
plt.subplot(3, 1, 3)
plt.plot(modeled_pars[:, 0][subset_array], ssd[subset_array], 'o')
plt.show()
def create_tacs(input_tac, times):
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac_smoothed, input_tac_splines = express.spline_interpolation(input_tac, times, new_time_steps)
pars_subset = express.combine_pars(bv=np.arange(0.1,0.60,0.1),
a=np.arange(1,15,0.5),
loc=np.arange(0.1,13,0.3),
scale=np.arange(0.1,6,0.3))
tacs = express.calc_tissue_tacs_mrx_3gamma(input_tac_smoothed, params = pars_subset, time_steps = new_time_steps, time_subset = times)
return tacs,pars_subset
def get_limit(self):
#load input tacs
times = self.rois.get_time_list()
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
input_tac_smoothed, input_tac_splines = express.spline_interpolation(
input_tac, times, new_time_steps)
try:
input_tac2 = np.array(self.rois.get_concentrations('porta'))
input_tac2 -= input_tac2[0]
input_tac_smoothed2, input_tac_splines2 = express.spline_interpolation(
input_tac2, times, new_time_steps)
except:
input_tac2 = None
pass
pars_subset = express.perf_pars_gen(bvs=np.arange(0.01, 0.1, 0.05),
mtts=np.arange(10, 100, 10),
sigmas=np.arange(0.1, 2, 0.5),
lags=(0, ))
vol4d = express.calc_tissue_tac_from_pars(input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=times,
params=pars_subset)
max_hu = np.max(vol4d, axis=0)
print np.array(np.where(pars_subset[:, 5] < 0.1)).shape
sp = np.array(
np.where(pars_subset[:, 1] < 0.1) +
np.where(pars_subset[:, 5] < 0.1))[0]
print sp.shape
plt.plot(times, vol4d[:, sp], 'o-', alpha=0.05, color='red')
plt.show()
def get_limit(self):
#load input tacs
times = self.rois.get_time_list()
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac = np.array(self.rois.get_concentrations('aorta'))
input_tac -= input_tac[0]
input_tac_smoothed, input_tac_splines = express.spline_interpolation(input_tac, times, new_time_steps)
try:
input_tac2 = np.array(self.rois.get_concentrations('porta'))
input_tac2 -= input_tac2[0]
input_tac_smoothed2, input_tac_splines2 = express.spline_interpolation(input_tac2, times, new_time_steps)
except:
input_tac2 = None
pass
pars_subset = express.perf_pars_gen(bvs=np.arange(0.01,0.1,0.05),
mtts=np.arange(10,100,10),
sigmas=np.arange(0.1,2,0.5),
lags=(0,))
vol4d = express.calc_tissue_tac_from_pars(input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=times,
params=pars_subset)
max_hu = np.max(vol4d ,axis=0)
print np.array(np.where(pars_subset[:,5]<0.1)).shape
sp = np.array(np.where(pars_subset[:,1]<0.1)+np.where(pars_subset[:,5]<0.1))[0]
print sp.shape
plt.plot(times,vol4d[:,sp],'o-',alpha=0.05,color='red')
plt.show()
tacs_real[l[0]]['value'].append(l[3])
else:
tacs_real[l[0]] = {'times': [], 'value': []}
print tacs_real
sample_size = len(tacs_real.keys())
pancreatic_pars = {
'a': np.random.normal(2, 0.5, size=sample_size),
'loc': np.random.normal(1, 0.5, size=sample_size),
'scale': np.random.normal(14, 5, size=sample_size),
'bv': np.random.normal(40, 15, size=sample_size)
}
#make input tac
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac_smoothed, input_tac_splines = express.spline_interpolation(
AORTA, times, new_time_steps)
time_steps = np.arange(10, 60, 2)
#lag and scale input
input_tac_smoothed = lag_input(input_tac_smoothed)
input_tac_smoothed -= input_tac_smoothed[0]
pars_subset = express.perf_pars_gen(bvs=np.arange(0.01, 0.7, 0.1),
mtts=np.arange(2, 40, 5),
sigmas=np.arange(0.1, 20, 10),
lags=(0, ),
bvs2=(0, ),
mtts2=(0, ),
sigmas2=(0, ),
lags2=(0, ))
pancreatic_pars = np.array([[1, 0.46, 0, 20, 0, 0, 0, 0],
[9, 0.28, 0, 16, 0, 0, 0, 0],
26,
29,
31,
33,
42,
47,
52,
57,
62,
91,
101,
111])
"""
#make input tac
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac_smoothed, input_tac_splines = express.spline_interpolation(AORTA, times, new_time_steps)
time_steps = np.arange(10,60,2)
#lag and scale input
input_tac_smoothed = lag_input(input_tac_smoothed)
input_tac_smoothed-=input_tac_smoothed[0]
pars_subset = express.perf_pars_gen(bvs=np.arange(0.01,0.7,0.1),
mtts=np.arange(2,40,5),
sigmas=np.arange(0.1,20,10),
lags=(0,),
bvs2=(0,),
mtts2=(0,),
sigmas2=(0,),
lags2=(0,))
pancreatic_pars = np.array([[1.2,0.7,0,81,0,0,0,0],
[9,0.28,0,16,0,0,0,0],
input_tac2 = np.array(input_tac2)-input_tac2[0]
#
BV_range_a = np.arange(0.01,0.99,0.4)
BV_range_p = (0.01,1,0.1)
MTT_range_a = np.arange(0.1,100,30)
MTT_range_p = (0,100,1)
S_range_a = np.arange(0.01,2,0.5)
S_range_p = (0.01,2,0.1)
#Simulation
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac_smoothed, input_tac_splines = expr.spline_interpolation(input_tac, times, new_time_steps)
input_tac_smoothed2, input_tac_splines2 = expr.spline_interpolation(input_tac2, times, new_time_steps)
input_tac_smoothed2[new_time_steps < times[0]] = input_tac2[0]
input_tac_smoothed[new_time_steps < times[0]] = input_tac[0]
reference_tacs, reference_pars = expr.calc_tissue_tac_mrx_conv2(input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=times,
rc_type='lognorm',
mtts=MTT_range_a,
bvs=BV_range_a,
rc_sigma=S_range_a,
lags=[0])
input_tac2 = np.array(input_tac2) - input_tac2[0]
#
BV_range_a = np.arange(0.01, 0.99, 0.4)
BV_range_p = (0.01, 1, 0.1)
MTT_range_a = np.arange(0.1, 100, 30)
MTT_range_p = (0, 100, 1)
S_range_a = np.arange(0.01, 2, 0.5)
S_range_p = (0.01, 2, 0.1)
#Simulation
new_time_steps = np.arange(times[0], times[-1] + 1, 1)
input_tac_smoothed, input_tac_splines = expr.spline_interpolation(
input_tac, times, new_time_steps)
input_tac_smoothed2, input_tac_splines2 = expr.spline_interpolation(
input_tac2, times, new_time_steps)
input_tac_smoothed2[new_time_steps < times[0]] = input_tac2[0]
input_tac_smoothed[new_time_steps < times[0]] = input_tac[0]
reference_tacs, reference_pars = expr.calc_tissue_tac_mrx_conv2(
input_tac_smoothed,
input_tac_smoothed2,
time_steps=new_time_steps,
time_subset=times,
rc_type='lognorm',
mtts=MTT_range_a,
bvs=BV_range_a,
rc_sigma=S_range_a,